Graphics processing units (GPUs) are all the rage these days and have helped to make Nvidia one of, if not the, most valuable company in the world, depending on what day of the week it is. But will GPUs become the norm and be at the heart of all future network platforms, managing multiple workloads including those related to communications networking and services?
This was a topic that came up for debate during the recent DSP Leaders World Forum session titled Building digital infrastructure from core to edge. The session, which was co-hosted by Vodafone UK’s head of infrastructure and energy, Francesca Serravalle, examined a number of trends and topics related to the nature of cloud-oriented next-generation telecom networks that can support multiple services, the role of core platforms, and much more. Serravalle was joined by fellow speakers from Rakuten Symphony, Verizon Business and Wind River.
To accompany the discussion, we set up a poll for our in-person forum attendees and the virtual audience that watched the sessions live via TelecomTV’s streaming platform. We asked: Will all future network platforms need to be GPU-based? Each person who voted could only choose one of three options.
As you can see from the chart above, only 20% of our respondents believed GPUs will be at the heart of everything, while a slight majority, 52%, were sure they will not, and 28% were not sure either way.
The topic came to the fore again this week when test and measurement vendor Rohde & Schwarz announced it had joined the AI-RAN Alliance, an industry body formed earlier this year (with Nvidia front and centre) that believes future radio access network (RAN) architectures need to be revamped to enable multiple, demanding workloads to run on a single platform – see AI-RAN Alliance launches at #MWC24.
Check out the full forum discussion on demand here.
– Ray Le Maistre, Editorial Director, TelecomTV